SBOU246 January   2022 TMP61 , TMP61-Q1 , TMP63 , TMP63-Q1 , TMP64 , TMP64-Q1

 

  1.   Trademarks
  2. 1Introduction
    1. 1.1 NTC Thermistor Versus TMP6 Linear Thermistor Family
    2. 1.2 NTC/Linear Thermistor TCR
    3. 1.3 NTC Versus Silicon-Based Linear Thermistor Trade-Offs
    4. 1.4 TMP6 Accuracy
  3. 2Typical NTC Thermistor Design Considerations
    1. 2.1 Voltage-Biased NTC Thermistor Network
    2. 2.2 Pinouts/Polarity
    3. 2.3 Converting NTC Thermistor Hardware Design to TMP6 Linear Thermistor Design
    4. 2.4 Simple Look-Up Table
  4. 3Software Changes
    1. 3.1 Firmware Design Considerations
    2. 3.2 Oversampling
    3. 3.3 Low-Pass Filtering in HW Versus SW
    4. 3.4 Calibration
  5. 4Design considerations for Full-Scale Range Voltage Output
    1. 4.1 Simple Current-Biased
    2. 4.2 Active Voltage-Biased
  6. 5Conclusion
  7. 6Additional Resources/Considerations
    1. 6.1 Constant-Current Source Design
    2. 6.2 TMP6 Thermistor Standard Component Footprints
    3. 6.3 Dual-Sourcing Approach for TMP6 and NTC Thermistors

3 Software Changes

When the output of the thermistor circuit is sensed by an ADC and converted to digital information for the MCU to process, the output must be converted to a temperature value. One of the common software R-T conversion methods is using a look-up table. This involves pre-populating a table of resistances and the associated temperature value for those resistances. The code will determine which resistance value more closely aligns with the expected temperature value by interpolating between the points. This method results in a very simple setup for the R-T table but puts high demand on the flash memory requirements for the MCU and a lengthy array-parsing program. It can also be inaccurate due to system errors such as tolerance variations and temperature coefficients which may lead to divergence from the ideal R-T table.

A second method of temperature conversion that can save on memory is the Steinhart-Hart equation below. The equation can be implemented into the temperature sensing code for mapping to the R-T curve of the thermistor:

Equation 2. 1/T = A + B × ln(R) + C × ln3(R)

where T = temperature (in Kelvin); R = measured resistance value; A,B,C are the calculated coefficients.

With the use of the TMP6 linear thermistor, however, a much better algorithm is available for converting is the 4th order polynomial regression model below:

Equation 3. T = A4 × R4 + A3 × R3 + A2 × R2 + A1 × R + A0

where T = temperature (in Celsius); R = measured resistance value; A0-4 are the calculated polynomial coefficients.

This approximation works well because of the linearity of the device and does not work for a non-linear NTC thermistor. The polynomial coefficients for the polynomial regression model can be generated in the Thermistor Design Tool.